On Fri, Oct 29, 2021 at 06:46:10AM -0700, Jakub Kicinski wrote: > On Fri, 29 Oct 2021 09:13:24 -0300 Jason Gunthorpe wrote: > > Jakub's path would be to test vlan_dev->reg_state != NETREG_REGISTERED > > in the work queue, but that feels pretty hacky to me as the main point > > of the UNREGISTERING state is to keep the object alive enough that > > those with outstanding gets can compelte their work and release the > > get. Leaving a wrecked object in UNREGISTERING is a bad design. > > That or we should investigate if we could hold the ref for real_dev all > the way until vlan_dev_free(). The latter is certainly better if it works out, no circular deps, etc. Thanks, Jason
